Path: utzoo!attcan!uunet!peregrine!elroy.jpl.nasa.gov!swrinde!zaphod.mps.ohio-state.edu!wuarchive!udel!haven!mimsy!drinkme.cs.umd.edu!kohout
From: kohout@drinkme.cs.umd.edu (Robert Kohout)
Newsgroups: comp.ai.philosophy
Subject: Re: Searle, Strong AI, and Chinese Rooms
Message-ID: <28345@mimsy.umd.edu>
Date: 5 Dec 90 19:26:42 GMT
References: <1990Nov15.204949.12075@Solbourne.COM> <1990Nov19.191925.28285@cs.umn.edu> <15798@venera.isi.edu> <1990Nov30.231103.17041@cs.umn.edu> <15878@venera.isi.edu>
Sender: news@mimsy.umd.edu
Reply-To: kohout@drinkme.cs.umd.edu (Robert Kohout)
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 48

In article <15878@venera.isi.edu> smoliar@vaxa.isi.edu (Stephen Smoliar) writes:
>Thus, having told his story about the imitation game, Turing still dismisses
>the prospect of pondering his original question as basically a waste of time.
>As I have said before, I am inclined to agree with him, leaving the question
>to philosophers while the engineers go off and try to do something useful.
>

I agree wholeheartedly, especially insofar as it applies to this Chinese
Room business. Leaving aside for a moment my objections to his "proof"
let us accept Searle's gedanken experiment as valid. Of what practical
importance is it? He is telling us that, if you ever build a machine
that can pass the Turing Test, it still won't "think". This sounds
like an issue to be debated in Star Trek, The Next Generation, not
here, not now.

If, on the other hand, Searle were trying to show that digital systems
alone will not be capable of passing the Turing Test, I would be much
more concerned. This is, I believe, one of Steve Harnad's basic
tenets. "Intelligence", whatever it may be, may simply not be computable.

I know that some of you will want to object: "But the brain in just
a big finite state machine". That is a conjecture, and for that matter
an old one which has been largely discredited. Neurons are cells,
real analog devices with certain behaviors that can be characterized
as digital behaviors. It is by no means certain that they are, in fact,
strictly digital in nature.

Some of you may even want to go further. That is, you may say
"At the fundamental level, matter is discreet, so at least in theory
we should be able to model the way it behaves." Again, this is
incorrect, on two counts. First, modern physics is somewhat confused
about the nature of "fundamental" particles: they behave as both waves
and particles. They are discreet, but also analog. Secondly, theory
also tells us that we will never (as in NEVER) be able to model the
world, or even a single brain, at the level of fundamental particles.
So we might as well adopt a sort of Heisenberg's uncertainly principle
and assume that, theoretically, we cannot model the behavior of
a complex material system at the level of fundamental particles.

Simple objections aside, the question remains: is "intelligence" 
computable? We immediately face the problem of having to define
intelligence. To simplify then, and bring this back 'round to Searle:
is it possible for a digital computer to pass the linguistic Turing
Test? That is, can the Chinese Room itself ever be built? Now
THAT is to me a more substantive and pertinent question that needs
to be address, by philosopher and engineer alike.

		- Bob Kohout